Merge branch 'pgo-estimated-trip-count' into fix-peel-branch-weights

Author: jdenny-ornl

.ci/all_requirements.txt

@@ -194,9 +194,9 @@ ml-dtypes==0.5.1 ; python_version < "3.13" \
     --hash=sha256:d13755f8e8445b3870114e5b6240facaa7cb0c3361e54beba3e07fa912a6e12b \
     --hash=sha256:fd918d4e6a4e0c110e2e05be7a7814d10dc1b95872accbf6512b80a109b71ae1
     # via -r mlir/python/requirements.txt
-nanobind==2.7.0 \
-    --hash=sha256:73b12d0e751d140d6c1bf4b215e18818a8debfdb374f08dc3776ad208d808e74 \
-    --hash=sha256:f9f1b160580c50dcf37b6495a0fd5ec61dc0d95dae5f8004f87dd9ad7eb46b34
+nanobind==2.9.2 \
+    --hash=sha256:c37957ffd5eac7eda349cff3622ecd32e5ee1244ecc912c99b5bc8188bafd16e \
+    --hash=sha256:e7608472de99d375759814cab3e2c94aba3f9ec80e62cfef8ced495ca5c27d6e
     # via -r mlir/python/requirements.txt
 numpy==2.0.2 \
     --hash=sha256:0123ffdaa88fa4ab64835dcbde75dcdf89c453c922f18dced6e27c90d1d0ec5a \
@@ -383,6 +383,10 @@ swig==4.3.1 \
     --hash=sha256:efec16327029f682f649a26da726bb0305be8800bd0f1fa3e81bf0769cf5b476 \
     --hash=sha256:fc496c0d600cf1bb2d91e28d3d6eae9c4301e5ea7a0dec5a4281b5efed4245a8
     # via -r lldb/test/requirements.txt
+typing-extensions==4.15.0 \
+    --hash=sha256:0cea48d173cc12fa28ecabc3b837ea3cf6f38c6d1136f85cbaaf598984861466 \
+    --hash=sha256:f0fa19c6845758ab08074a0cfa8b7aecb71c999ca73d62883bc25cc018c4e548
+    # via -r mlir/python/requirements.txt
 urllib3==2.5.0 \
     --hash=sha256:3fc47733c7e419d4bc3f6b3dc2b4f890bb743906a30d56ba4a5bfa4bbff92760 \
     --hash=sha256:e6b01673c0fa6a13e374b50871808eb3bf7046c4b125b216f6bf1cc604cff0dc

.ci/monolithic-windows.sh

@@ -56,6 +56,8 @@ start-group "ninja"
 # Targets are not escaped as they are passed as separate arguments.
 ninja -C "${BUILD_DIR}" -k 0 ${targets} |& tee ninja.log
 
-start-group "ninja runtimes"
-
-ninja -C "${BUILD_DIR}" -k 0 ${runtimes_targets} |& tee ninja_runtimes.log
+if [[ "${runtime_targets}" != "" ]]; then
+  start-group "ninja runtimes"
+  
+  ninja -C "${BUILD_DIR}" -k 0 ${runtimes_targets} |& tee ninja_runtimes.log
+fi

.github/CODEOWNERS

@@ -17,6 +17,7 @@
 /runtimes/ @llvm/reviewers-libcxx
 
 /llvm/lib/Analysis/BasicAliasAnalysis.cpp @nikic
+/llvm/lib/Analysis/HashRecognize.cpp @artagnon @pfusik
 /llvm/lib/Analysis/InstructionSimplify.cpp @nikic
 /llvm/lib/Analysis/LazyValueInfo.cpp @nikic
 /llvm/lib/Analysis/ScalarEvolution.cpp @nikic

bolt/lib/Core/BinaryContext.cpp

@@ -207,15 +207,15 @@ Expected<std::unique_ptr<BinaryContext>> BinaryContext::createBinaryContext(
                              Twine("BOLT-ERROR: ", Error));
 
   std::unique_ptr<const MCRegisterInfo> MRI(
-      TheTarget->createMCRegInfo(TripleName));
+      TheTarget->createMCRegInfo(TheTriple));
   if (!MRI)
     return createStringError(
         make_error_code(std::errc::not_supported),
         Twine("BOLT-ERROR: no register info for target ", TripleName));
 
   // Set up disassembler.
   std::unique_ptr<MCAsmInfo> AsmInfo(
-      TheTarget->createMCAsmInfo(*MRI, TripleName, MCTargetOptions()));
+      TheTarget->createMCAsmInfo(*MRI, TheTriple, MCTargetOptions()));
   if (!AsmInfo)
     return createStringError(
         make_error_code(std::errc::not_supported),
@@ -227,7 +227,7 @@ Expected<std::unique_ptr<BinaryContext>> BinaryContext::createBinaryContext(
   AsmInfo->setAllowAtInName(true);
 
   std::unique_ptr<const MCSubtargetInfo> STI(
-      TheTarget->createMCSubtargetInfo(TripleName, "", FeaturesStr));
+      TheTarget->createMCSubtargetInfo(TheTriple, "", FeaturesStr));
   if (!STI)
     return createStringError(
         make_error_code(std::errc::not_supported),

bolt/lib/Target/AArch64/AArch64MCPlusBuilder.cpp

@@ -2517,10 +2517,8 @@ class AArch64MCPlusBuilder : public MCPlusBuilder {
   createInstrIncMemory(const MCSymbol *Target, MCContext *Ctx, bool IsLeaf,
                        unsigned CodePointerSize) const override {
     unsigned int I = 0;
-    InstructionListType Instrs(IsLeaf ? 12 : 10);
+    InstructionListType Instrs(10);
 
-    if (IsLeaf)
-      createStackPointerIncrement(Instrs[I++], 128);
     createPushRegisters(Instrs[I++], AArch64::X0, AArch64::X1);
     getSystemFlag(Instrs[I++], AArch64::X1);
     InstructionListType Addr = materializeAddress(Target, Ctx, AArch64::X0);
@@ -2535,8 +2533,6 @@ class AArch64MCPlusBuilder : public MCPlusBuilder {
     loadReg(Instrs[I++], AArch64::X2, AArch64::SP);
     setSystemFlag(Instrs[I++], AArch64::X1);
     createPopRegisters(Instrs[I++], AArch64::X0, AArch64::X1);
-    if (IsLeaf)
-      createStackPointerDecrement(Instrs[I++], 128);
     return Instrs;
   }
 

bolt/runtime/instr.cpp

@@ -714,9 +714,11 @@ static char *getBinaryPath() {
       uint32_t Ret = __readlink(FindBuf, TargetPath, sizeof(TargetPath));
       assert(Ret != -1 && Ret != BufSize, "readlink error");
       TargetPath[Ret] = '\0';
+      __close(FDdir);
       return TargetPath;
     }
   }
+  __close(FDdir);
   return nullptr;
 }
 

bolt/test/link_fdata.py

@@ -32,7 +32,7 @@
 
 # Regexes to extract FDATA lines from input and parse FDATA and pre-aggregated
 # profile data
-prefix_pat = re.compile(f"^# {args.prefix}: (.*)")
+prefix_pat = re.compile(f"^(#|//) {args.prefix}: (.*)")
 
 # FDATA records:
 # <is symbol?> <closest elf symbol or DSO name> <relative FROM address>
@@ -61,7 +61,7 @@
         prefix_match = prefix_pat.match(line)
         if not prefix_match:
             continue
-        profile_line = prefix_match.group(1)
+        profile_line = prefix_match.group(2)
         fdata_match = fdata_pat.match(profile_line)
         preagg_match = preagg_pat.match(profile_line)
         nolbr_match = nolbr_pat.match(profile_line)

bolt/test/runtime/copy_file.py

@@ -0,0 +1,15 @@
+import sys
+import shutil
+
+with open(sys.argv[1] + ".output") as log_file:
+    lines = log_file.readlines()
+    for line in lines:
+        if line.startswith(sys.argv[2]):
+            pid = line.split(" ")[1].strip()
+            shutil.copy(
+                sys.argv[1] + "." + pid + ".fdata",
+                sys.argv[1] + "." + sys.argv[3] + ".fdata",
+            )
+            sys.exit(0)
+
+sys.exit(1)

bolt/test/runtime/instrumentation-indirect-2.c

@@ -50,7 +50,7 @@ int main() {
   return 0;
 }
 /*
-REQUIRES: system-linux,shell,fuser
+REQUIRES: system-linux,fuser
 
 RUN: %clang %cflags %s -o %t.exe -Wl,-q -pie -fpie
 
@@ -61,10 +61,14 @@ RUN: --instrumentation-wait-forks
 
 # Instrumented program needs to finish returning zero
 # Both output and profile must contain all 16 functions
-RUN: %t.instrumented_conservative > %t.output
-# Wait for profile and output to be fully written
-RUN: bash %S/wait_file.sh %t.output
-RUN: bash %S/wait_file.sh %t.fdata
+# We need to use bash to invoke this as otherwise we hang inside a
+# popen.communicate call in lit's internal shell. Eventually we should not
+# need this.
+# TODO(boomanaiden154): Remove once
+# https://github.com/llvm/llvm-project/issues/156484 is fixed.
+RUN: bash -c "%t.instrumented_conservative; wait" > %t.output
+# We can just read because we ensure the profile will be fully written by
+# calling wait inside the bash invocation.
 RUN: cat %t.output | FileCheck %s --check-prefix=CHECK-OUTPUT
 RUN: cat %t.fdata | FileCheck %s --check-prefix=CHECK-COMMON-PROF
 
@@ -112,14 +116,8 @@ RUN: bash %S/wait_file.sh %t.output
 # Make sure all functions were called
 RUN: cat %t.output | FileCheck %s --check-prefix=CHECK-OUTPUT
 
-RUN: child_pid=$(cat %t.output | grep funcA | awk '{print $2;}')
-RUN: par_pid=$(cat %t.output | grep funcB | awk '{print $2;}')
-
-RUN: bash %S/wait_file.sh %t.$child_pid.fdata
-RUN: bash %S/wait_file.sh %t.$par_pid.fdata
-
-RUN: mv %t.$child_pid.fdata %t.child.fdata
-RUN: mv %t.$par_pid.fdata %t.parent.fdata
+RUN: %python %S/copy_file.py %t funcA child
+RUN: %python %S/copy_file.py %t funcB parent
 
 # Instrumented binary must produce two profiles with only local calls
 # recorded. Functions called only in child should not appear in parent's

bolt/test/timers.c

@@ -1,22 +1,21 @@
-/* This test checks timers for metadata manager phases.
-# RUN: %clang %cflags %s -o %t.exe
-# RUN: link_fdata %s %t.exe %t.fdata
-# RUN: llvm-bolt %t.exe -o %t.null --data %t.fdata -w %t.yaml --time-rewrite \
-# RUN:   2>&1 | FileCheck %s
-# RUN: link_fdata %s %t.exe %t.preagg PREAGG
-# RUN: perf2bolt %t.exe -o %t.null -p %t.preagg --pa --time-rewrite \
-# RUN:   2>&1 | FileCheck %s --check-prefix=CHECK-P2B
+// This test checks timers for metadata manager phases.
+// RUN: %clang %cflags %s -o %t.exe
+// RUN: link_fdata %s %t.exe %t.fdata
+// RUN: llvm-bolt %t.exe -o %t.null --data %t.fdata -w %t.yaml --time-rewrite \
+// RUN:   2>&1 | FileCheck %s
+// RUN: link_fdata %s %t.exe %t.preagg PREAGG
+// RUN: perf2bolt %t.exe -o %t.null -p %t.preagg --pa --time-rewrite \
+// RUN:   2>&1 | FileCheck %s --check-prefix=CHECK-P2B
 
-# CHECK-DAG: update metadata post-emit
-# CHECK-DAG: process section metadata
-# CHECK-DAG: process metadata pre-CFG
-# CHECK-DAG: process metadata post-CFG
-# CHECK-DAG: finalize metadata pre-emit
+// CHECK-DAG: update metadata post-emit
+// CHECK-DAG: process section metadata
+// CHECK-DAG: process metadata pre-CFG
+// CHECK-DAG: process metadata post-CFG
+// CHECK-DAG: finalize metadata pre-emit
 
-# CHECK-P2B-DAG: process section metadata
-# CHECK-P2B-DAG: process metadata pre-CFG
+// CHECK-P2B-DAG: process section metadata
+// CHECK-P2B-DAG: process metadata pre-CFG
 
-# FDATA: 0 [unknown] 0 1 main 0 1 0
-# PREAGG: B X:0 #main# 1 0
-*/
+// FDATA: 0 [unknown] 0 1 main 0 1 0
+// PREAGG: B X:0 #main# 1 0
 int main() { return 0; }